This dir contains the C programs of ANFIS architecture proposed in [1].

Author: Jyh-Shing Roger Jang

Organization: Dept of EECS, UC Berkeley

Subdirectory information:
	anfis3: main program of anfis, with batch learning
	c_lib: handy C functions
	common: C functions called by main program of anfis
	example1: simulation example 1 of [1].
	example2: simulation example 2 of [1].
	example3: simulation example 3 of [1].
	example4: simulation example 4 of [1].
	misc: misc C programs for preprocessing data files
	test: for test only

How to run it for each example:
	(1) Enter anfis3 and type "make" to compile the program
	(2) Enter example1, say.
	(3) If there is no 'config' file, generate it by
            the command misc/gen_conf. 
	(4) If there is no initial parameter file 'para.ini', generate it
            by the command misc/initpara. 
	(5) Modify the script go3 if necessary.
	(6) Type "go3" to run the program.

File convention in each example dir:
	config: incidence matrix showing the connnction of the adaptive network
	data.trn: training data file
	data.chk: checking (test or validating) data file
	para.ini: initial parameter file
	para.***: parameter file after *** epochs (*** is a number)
	para.fin: final parameter file
	error.trn: training RMSE error file
	error.chk: checking RMSE error file
	go3: command shell script

Extension:
	Extension of this program to the backpropagation through time is
	reported in [3]. It should be straightforward to modify this
	program to serve that purpose. Extension to sequential learning, as
	proposed in [1], should be straightforward, too.

Reference:
	[1] "ANFIS: Adaptive-Network-based Fuzzy Inference Systems", IEEE
	    Trans. on Systems, Man and Cybernetics, May 1993.
	[2] "Functional Equivalence between Radial Basis Function Networks and
	    Fuzzy Inference Systems", Trans. on NN, Jan. 1993.
	[3] "Self-Learning Fuzzy Controllers based on Temporal
	    Back-Propagation", Trans. on NN, Sept 1992.

Disclaimer:
	This program is provided as it is.
	The author is not responsible for any damage, directly or indirectly,
        caused by this program.

=======================================================================
Jyh-Shing Roger Jang                    Research Associate
571 Evans, Dept. of EECS		Phone: 510-642-5029
University of California 		Fax  : 510-642-5775
Berkeley, CA 94720			Email: jang@eecs.berkeley.edu
=======================================================================
